← Interview Prep

IP Addressing & CIDR — Subnetting & Planning

The addressing toolkit: CIDR notation & the mask, subnet math (hosts, block size, network/broadcast), the prefix table to memorize, RFC 1918 & special ranges, VLSM, route summarization/aggregation, wildcard masks, and network-planning principles.

Reading a prefix, subnetting a block on a whiteboard, and summarizing routes are table-stakes networking skills — and constant interview fodder. This is the addressing and planning toolkit: CIDR notation, the subnet math, the special ranges, VLSM, and aggregation, with the shortcuts that make it fast.

An IPv4 address is 32 bits split into a network part and a host part. The prefix length /N says how many leading bits are network. Everything else — host counts, the broadcast address, summarization — falls out of that one split.

CIDR & the mask

CIDR (Classless Inter-Domain Routing) dropped the old A/B/C classes: a prefix is just an address + a length. The mask is the prefix in dotted form — /24 = 255.255.255.0 (24 one-bits). Bits set = network; bits clear = host.

• Hosts in a /N = 2^(32−N) − 2 — subtract 2 for the network (all host bits 0) and broadcast (all host bits 1) addresses.
• Block size (addresses per subnet) = 2^(32−N) = 256 − mask-octet in the "interesting" octet — the fast mental trick.
• Network address = address AND mask; broadcast = network OR (inverted mask); usable range sits between them; the gateway is a convention (often .1), not a rule.

The table to memorize

*/31 is special: RFC 3021 lets a point-to-point link use both addresses (no network/broadcast waste). /32 is a single host — loopbacks, host routes, anycast VIPs.

Special & reserved ranges

VLSM — right-size each subnet

Variable-Length Subnet Masking = carve one block into subnets of different sizes to fit each need instead of wasting a /24 on a 2-host link. Allocate largest-first, on power-of-two boundaries.

Given 10.1.0.0/24, need: 100 hosts, 50 hosts, 10 hosts, 2 P2P links

10.1.0.0/25   (.0–.127)   → 126 hosts   [the 100-host LAN]
10.1.0.128/26 (.128–.191) → 62 hosts    [the 50-host LAN]
10.1.0.192/28 (.192–.207) → 14 hosts    [the 10-host LAN]
10.1.0.208/30 (.208–.211) → 2 hosts     [P2P #1]
10.1.0.212/30 (.212–.215) → 2 hosts     [P2P #2]   ...room left to grow

Aggregation / summarization

The inverse: combine contiguous subnets into one shorter prefix to shrink routing tables. Find the common leading bits; that's the summary length.

10.1.0.0/24  +  10.1.1.0/24  +  10.1.2.0/24  +  10.1.3.0/24
first 22 bits are identical  →  summary = 10.1.0.0/22   (covers .0–.3)

• Only contiguous, power-of-two-aligned blocks summarize cleanly — the reason you plan addressing hierarchically (allocate by site/role so each site is one summarizable block).
• Summarization + longest-prefix match is why a more specific route still wins over the summary when needed (see route selection).
• Supernet = a prefix shorter than the classful boundary (a summary); subnet = longer. Wildcard mask (ACLs/OSPF) is the bit-inverse of the subnet mask (0.0.0.255 = a /24).

Planning principles

• Hierarchical & contiguous: allocate a block per region/site/role so routes summarize upward — smaller tables, faster convergence, cleaner filtering.
• Leave room to grow and align on power-of-two boundaries (don't hand out odd sizes).
• /31 for P2P, /32 for loopbacks; reserve a consistent gateway (.1).
• IPv6: plan on nibble boundaries — /64 per LAN (always), /56 or /48 per site; think in prefixes, not host counts. See IPv6.
• Automate the math — the same operations appear as code in Python — Ranges & Parsing.